U.S. patent application number 14/646360 was filed with the patent office on 2015-10-29 for apparatus and methods for estimating optical ethernet data sequences.
The applicant listed for this patent is Yin HUANG, Jian LI, QUALCOMM INCORPORATED, Yisheng XUE, Haitao ZHANG. Invention is credited to Yin HUANG, Jian LI, Yisheng XUE, Haitao ZHANG.
Application Number | 20150312056 14/646360 |
Document ID | / |
Family ID | 51166470 |
Filed Date | 2015-10-29 |
United States Patent
Application |
20150312056 |
Kind Code |
A1 |
ZHANG; Haitao ; et
al. |
October 29, 2015 |
APPARATUS AND METHODS FOR ESTIMATING OPTICAL ETHERNET DATA
SEQUENCES
Abstract
A receiver is disclosed that includes a slicer having an input
to receive a sequence of symbols exhibiting
inter-symbol-interference (ISI). The slicer determines a state
associated with each symbol based on a threshold. A feedback
equalization unit is coupled to the slicer to apply equalization to
the symbol fed to the slicer input based on prior detected symbol
states. A Least-Mean-Square (LMS) unit cooperates with the slicer
and feedback equalization unit to estimate a channel impulse
response based on the equalized symbols. The LMS unit feeds the
estimated channel impulse response to a
Maximum-Likelihood-Sequence-Estimation (MLSE) unit to generate an
estimated sequence of bits based on the estimated channel impulse
response.
Inventors: |
ZHANG; Haitao; (Beijing,
CN) ; HUANG; Yin; (Beijing, CN) ; LI;
Jian; (Beijing, CN) ; XUE; Yisheng; (Beijing,
CN) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
ZHANG; Haitao
HUANG; Yin
LI; Jian
XUE; Yisheng
QUALCOMM INCORPORATED |
San Diego
San Diego
San Diego
San Diego
San Diego |
CA
CA
CA
CA
CA |
US
US
US
US
US |
|
|
Family ID: |
51166470 |
Appl. No.: |
14/646360 |
Filed: |
January 8, 2013 |
PCT Filed: |
January 8, 2013 |
PCT NO: |
PCT/CN2013/070196 |
371 Date: |
May 20, 2015 |
Current U.S.
Class: |
375/233 |
Current CPC
Class: |
H04L 1/0054 20130101;
H04L 25/03057 20130101; H04B 10/60 20130101; H04L 25/0212
20130101 |
International
Class: |
H04L 25/02 20060101
H04L025/02; H04L 25/03 20060101 H04L025/03; H04L 1/00 20060101
H04L001/00; H04B 10/60 20060101 H04B010/60 |
Claims
1. A receiver comprising: a slicer having an input to receive a
sequence of symbols exhibiting inter-symbol-interference (ISI), the
slicer configured to detect a state associated with each symbol
based on a threshold; a feedback equalization unit coupled to the
slicer to apply equalization to the sequence of symbols based on
prior detected symbol states; a Least-Mean-Square (LMS) unit to
estimate a channel impulse response based on the equalized symbols;
and a Maximum-Likelihood-Sequence-Estimation (MLSE) unit to
generate an estimated sequence of symbols based on the estimated
channel impulse response.
2. The receiver of claim 1 embodied as a 10GBASE-LRM Ethernet
receiver.
3. The receiver of claim 1, wherein the feedback equalization unit
includes a selectable number of feedback paths coupled to
respective equalizer taps, a number of the selectable feedback
paths enabled based on a correlation matrix.
4. The receiver of claim 3, wherein the equalizer taps are shared
with the LMS unit.
5. The receiver of claim 1, wherein the estimated channel impulse
response generated by the LMS unit comprises a coarse estimation of
a channel impulse response, and the MLSE unit generates a fine
estimation of the channel impulse response based on the coarse
estimation.
6. The receiver of claim 5, wherein the MLSE unit cooperates with
the LMS unit to generate the fine estimation of the channel impulse
response.
7. The receiver of claim 1, wherein the feedback equalization unit
includes plural selectable feedback paths, and wherein the feedback
paths are activated in an incremental manner.
8. The receiver of claim 1, wherein the feedback equalization unit
includes plural selectable feedback paths, and wherein the feedback
paths are activated simultaneously.
9. The receiver of claim 1, wherein the feedback equalization unit
includes plural selectable feedback paths including a first set of
feedback paths, and a second set of feedback paths that forms a
replica of the first set of feedback paths.
10. The receiver of claim 9, wherein the LMS unit is shared by the
first and second sets of feedback paths.
11. The receiver of claim 1, wherein in a first mode, the estimated
sequence comprises a coarse estimation of a channel impulse
response.
12. The receiver of claim 11, wherein in a second mode, the
estimated sequence comprises a fine estimation of the channel
impulse response.
13. The receiver of claim 1, further comprising a linear equalizer
coupled to the input of the slicer.
14. The receiver of claim 13, wherein the linear equalizer
comprises a feed-forward equalizer having fixed taps to compensate
for pre-cursor ISI.
15. A method of determining an output data sequence from an input
sequence affected by ISI, the method comprising: receiving an input
symbol sequence; slicing the input symbol sequence to initially
determine respective symbol states; feeding back selected
determined symbol states to equalize the input sequence; estimating
a channel impulse response based on the slicing and feeding back;
and estimating a most likely sequence of output bits based on the
estimated channel impulse response.
16. The method of claim 15, further comprising: iteratively
updating the estimated channel impulse response to track changes in
the channel impulse response.
17. The method of claim 15, wherein feeding back selected
determined symbol states comprises: incrementally activating plural
feedback paths.
18. The method of claim 15, wherein feeding back selected
determined symbol states comprises: simultaneously activating
plural feedback paths.
19. The method of claim 15, wherein estimating a channel impulse
response comprises: generating a coarse estimate of a channel
impulse response.
20. The method of claim 19, wherein estimating further comprises:
generating a fine estimate of the channel impulse response based on
the coarse estimate.
Description
TECHNICAL FIELD
[0001] The disclosure herein relates generally to high-speed
Ethernet systems and associated methods, and more particularly to
Long Range Multimode (LRM) implementations of 10GBASE
receivers.
BACKGROUND OF RELATED ART
[0002] A variety of high-speed Ethernet standards have been
proposed for transferring data at speeds up to 10 Gb/s. One such
standard, 10GBASE-LRM, transmits data over optical multimode fibers
(MMF) up to distances of 220 meters. Backwards compatibility
requirements for the 10GBASE-LRM standard dictate that the receiver
circuitry be able to decode data transmissions accurately without
an initial training sequence of data to calibrate the channel. As a
result, 10GBASE-LRM receivers employ electronic dispersion
compensation (EDC) technology to accurately resolve incoming data
streams that are susceptible to inter-symbol-interference
(ISI).
[0003] One conventional way to achieve suitable EDC results is to
employ a Viterbi-decoder that takes a known channel impulse
response (CIR) as an input and estimates an output sequence from an
ISI-impacted input sequence of bits. Typically, a feed-forward
equalizer (FFE) having complex tap adaptation and selection
circuitry processes an input bit sequence and cooperates with a
Least-Mean-Square Channel Estimator (LMS-CE) to generate the
channel impulse response. The CIR is fed to a
"Maximum-Likelihood-Sequence-Estimation" (MLSE) circuit, which
includes the Viterbi decoder.
[0004] While the conventional sequence estimation circuit described
above works well for its intended applications, the use of a
complex feed-forward equalizer generally involves a form of finite
impulse response filtering, which often results in multiplication
operations for a digital system processor (DSP). Further, the FFE
generally includes a set of adaptive taps that are separate from
the LMS-CE taps, thus adding to the circuit complexity.
BRIEF DESCRIPTION OF THE DRAWINGS
[0005] The present embodiments are illustrated by way of example
and are not intended to be limited by the figures of the
accompanying drawings.
[0006] FIG. 1 is a block diagram of one embodiment of a 10GBASE-LRM
channel.
[0007] FIG. 2 illustrates one embodiment of an EDC architecture
suitable for use in the channel of FIG. 1.
[0008] FIG. 3 shows one detailed implementation of a coarse channel
estimation circuit for use in an EDC architecture such as that
illustrated in FIG. 2.
[0009] FIG. 4 illustrates an alternative embodiment of a coarse
channel estimation circuit similar to the circuit shown in FIG.
3.
[0010] FIG. 5 illustrates a flowchart of steps to determine a
coarse channel estimation utilizing either of the coarse channel
estimation circuits of FIG. 3 or 4.
[0011] FIG. 6 illustrates steps employed in a method to determine a
fine channel estimation based on the coarse channel estimation
generated in the method steps of FIG. 5.
[0012] FIG. 7 illustrates a coarse estimate of a channel impulse
response using the steps of FIG. 5, and a fine estimate of a
channel impulse response using the steps of FIG. 6.
DETAILED DESCRIPTION
[0013] Embodiments of receiver circuits and associated methods are
disclosed. In one embodiment, the receiver circuit includes a
slicer having an input to receive a sequence of symbols exhibiting
inter-symbol-interference (ISI). The slicer determines a state
associated with each symbol based on a threshold input. Feedback
equalization circuitry is coupled to the slicer to apply
equalization to the symbol fed to the slicer input based on prior
detected symbol states. A Least-Mean-Square (LMS) circuit
cooperates with the slicer and feedback equalization circuitry to
estimate a channel impulse response based on the equalized symbols.
The LMS circuit feeds the estimated channel impulse response to a
Most-Likely-Sequence-Estimation (MLSE) circuit to generate an
estimated sequence of bits based on the estimated channel impulse
response. By implementing a slicer-based circuit to estimate the
channel impulse response instead of an FIR filter, the complexity
of the EDC circuitry may be reduced.
[0014] In one embodiment, a method of determining an output data
sequence from an input sequence affected by ISI is disclosed. The
method includes receiving an input symbol sequence and slicing the
input symbol sequence to initially determine respective symbol
states. The selected determined symbol states are fed back to the
slicer input to equalize the input symbol sequence. A channel
impulse response is estimated based on the slicing and feeding
back. A most likely sequence of output bits is then estimated based
on the estimated channel impulse response.
[0015] In the following description, numerous specific details are
set forth such as examples of specific components, circuits, and
processes to provide a thorough understanding of the present
disclosure. Also, in the following description and for purposes of
explanation, specific nomenclature is set forth to provide a
thorough understanding of the present embodiments. However, it will
be apparent to one skilled in the art that these specific details
may not be required to practice the present embodiments. In other
instances, well-known circuits and devices are shown in block
diagram form to avoid obscuring the present disclosure. The term
"coupled" as used herein means connected directly to or connected
through one or more intervening components or circuits. Any of the
signals provided over various buses described herein may be
time-multiplexed with other signals and provided over one or more
common buses. Additionally, the interconnection between circuit
elements or software blocks may be shown as buses or as single
signal lines. Each of the buses may alternatively be a single
signal line, and each of the single signal lines may alternatively
be buses, and a single line or bus might represent any one or more
of a myriad of physical or logical mechanisms for communication
between components. The present embodiments are not to be construed
as limited to specific examples described herein but rather to
include within their scopes all embodiments defined by the appended
claims.
[0016] More specifically, and referring generally to FIG. 1, one
embodiment of a 10GBASE-LRM channel 100 is illustrated. The
10GBASE-LRM channel 100 employs a transmitter optical sub-assembly
(TOSA) 102 to transmit data along multimode fiber (MMF) 105 to a
link partner receiver 104. The receiver 104 includes a receiver
optical sub-assembly (ROSA) 106 that employs a photodiode (PD) 110,
transimpedance amplifier (TIA) 112, and an optical-to-electrical
interface (I/F) 114 that couples to the TIA.
[0017] The ROSA 106 generally converts the optical signals
transmitted along the multimode fiber 105 to electrical signals.
The electrical signals are then fed to an electrical sub-assembly
(ESA) 108 that includes an electronic dispersion compensation unit
(EDC) 116. The EDC unit 116 estimates a channel impulse response
(CIR), and includes an MLSE unit 118 as more fully described below.
A digital signal processor (DSP) 120 interacts with the EDC 116 to
estimate the correct sequence of bits that were fed to the EDC 116
based on an estimation technique such as, for example, a
Viterbi-based algorithm.
[0018] FIG. 2 illustrates further detail of one embodiment of an
EDC, generally designated 200, that minimizes complexity within the
receiver 104 to thus enhance implementation. The EDC 200 includes a
counter 202 that monitors a count of the number of symbols received
along an input path 203, and provides a control signal based on the
count to a selector 204. The selector 204 steers the bit sequence
first to a coarse channel estimation unit 208. The coarse channel
estimation unit 208 generally develops a first estimation of the
channel response. In one embodiment, the coarse channel estimation
unit 208 includes a slicer with feedback 212. The slicer with
feedback 212 generally makes decisions of each symbol state. The
coarse channel estimation unit 208 also includes a
Least-Mean-Square Channel Estimator (LMS-CE) 214 that cooperates
with the slicer with feedback 212 to generate the estimated initial
channel response.
[0019] Further referring to FIG. 2, the initial channel response
generated by the coarse channel estimation unit 208 is fed to a
fine channel estimation unit 210. The fine channel estimation unit
generally develops a more accurate channel response for determining
the sequence of symbols. The fine channel estimation unit 210
includes a Most-Likely-Sequence-Estimator (MLSE) unit 216 that in
one embodiment operates in accordance with a Viterbi algorithm to
generate an estimated output bit sequence. The MLSE unit 216
interfaces with a second LMS-CE unit 218 to further refine the
channel estimation. In one embodiment, a single LMS-CE unit may be
shared by the coarse channel estimation unit 208 and the fine
channel estimation unit 210. Once the fine channel estimation unit
210 has properly converged on an accurate channel response
estimation, data sequences may be steered directly to the fine
channel estimation unit from the input path 203 via the selector
204.
[0020] FIG. 3 illustrates further detail of one embodiment of a
coarse channel estimation unit, generally designated 300, that
provides the functionality of the coarse channel estimation unit
208 of FIG. 2. The unit 300 includes an input symbol summer 302
that aggregates received symbol voltages at 301 with precursor and
post-cursor compensation voltages, at 303 and 305, respectively,
that are provided by an optional equalizer unit 310 and incremental
feedback units 304. A slicer 306, such as a straightforward
comparator having an input reference to "0" volts, generates
decisions on the states of the received input symbols. The
incremental feedback units 304 cooperate with an LMS-CE unit 308 to
provide a level of feedback equalization to the summer 302 and
compensate for post-cursor inter-symbol-interference (ISI)
affecting the channel. For one embodiment, the optional linear
equalizer 310, such as a fixed-tap feed-forward equalizer, may be
employed in a limited capacity to compensate for pre-cursor
ISI.
[0021] Further referring to FIG. 3, the incremental feedback units
304 include a series-coupled delay line of delay elements 312 that
couples to the output of the slicer 306. The delay elements 312
each exhibit a delay of one symbol time T, or unit interval. As a
given symbol {tilde over (y)}(n) is output from the slicer 306, it
propagates through each delay element. A sequence of symbols thus
proceeds through the delay line in a serially-shifted manner,
generally corresponding to previous received symbols. Each delay
element 312 includes a tap output 314 that feeds a multiplier 316.
Each multiplier receives an inverted version of a particular
symbol, or bit value, and applies a weighting value to the symbol.
The outputs from the multipliers 316 are fed to a feedback summer
317, which aggregates or sums all of the feedback compensation
voltages from the various feedback paths and feeds the aggregated
feedback voltage to the input summer 302. The weightings applied to
each multiplier 316 generally dictate how each subsequent symbol
will contribute to the overall compensation applied at the summer
302. Generally, the most recent symbols usually contribute the most
to ISI, and the weightings for those taps are often higher than
more latent symbols.
[0022] Further referring to FIG. 3, the LMS-CE unit 308 includes
LMS-CE logic 318 that determines the weightings to apply to the
multipliers 316 based on an error-minimization algorithm such as
the well-known LMS algorithm. Other algorithms that optimize
weightings based on detected errors may also be employed. The
LMS-CE logic 318 receives a version of the sliced symbols and
applies the LMS algorithm to the symbols. An optional normalization
process may be carried out by a normalization unit 319. The LMS-CE
unit further includes a plurality of equalizer taps 320. Each tap
generally exhibits an adjustable gain corresponding to the desired
weighting for that tap, and generally corresponds to a post-cursor.
Post-cursors are ISI effects caused by one or more prior received
symbols. A first post-cursor tap is identified by h(k), with
subsequent post-cursor taps identified by h(k+1), through
h(k+m-1).
[0023] As explained earlier, for one embodiment, the weightings for
each tap may be adjusted based on the LMS algorithm applied to the
received symbol sequence. The taps are selectively applied to the
multipliers 316 via a CE tap selector 322 that controls a plurality
of switches to enable selected taps. The selected taps define a
channel estimation window more fully described below. In one
embodiment, the switches are sequentially activated in a sequence
to incrementally apply feedback from multiple feedback paths. Over
a series of iterations, the LMS algorithm will enable the coarse
channel estimation unit 300 to converge to a desired coarse channel
response through iterative adjustments of the tap weightings.
[0024] With continued reference to FIG. 3, the optional linear
equalizer unit 310 in one embodiment takes the form of a
feed-forward equalizer to compensate for pre-cursor ISI. The linear
equalizer unit 310 includes a main cursor path 324 that coincides
with the input path of the received sequence of symbols. A first
multiplier 326 applies a fixed or pre-programmed main cursor tap
value a0 (often a value of "1") to each received symbol. A delay
element 328 is selectively coupled to the main cursor path 324 and
exhibits a delay T corresponding to the symbol interval time. A
second multiplier 330 receives a first pre-cursor tap value a1 to
apply to the delayed symbol. The current symbol y(n) and an
inverted version of the delayed symbol representing a previously
received symbol y(n-1) are then summed at the summing unit 302 to
compensate for the anticipated effects of pre-cursor ISI.
[0025] In operation, the tap length of the channel being estimated
is set to a value "L", and the tap length of the channel estimation
window is set to "2L." In a first stage of operations, the LMS-CE
logic calculates the tap weightings based on the following
expressions:
e(n)=y(n-L)-h.sup.T(n){tilde over (y)}(n)
{circumflex over (h)}(n+1)={circumflex over (h)}(n)+.mu.e(n){tilde
over (y)}(n)
where {tilde over (y)}(n) represents the output of the slicer, h(n)
represents a channel estimation of the LMS-CE, e(n) represents an
error signal, and .mu. represents a step size. In a second stage of
operations, the CE tap selector 322 selects the largest consecutive
taps within the channel estimation window based on a criteria such
as peak energy. The detected symbol after ISI cancellation is given
by:
y ~ ( n ) = sgn { y ( n ) - m ' = 1 m h ^ ( k + m ' ) y ~ ( n - m '
) } , m = 1 , , M ##EQU00001##
where: h(k) represents a channel estimation tap with the peak
energy, m represents the current feedback tap number used for ISI
cancellation, M represents a maximum feedback tap number, and sgn {
}represents a sign function which extracts the sign of a real
number. The initial states of the tap weightings are based on the
tap feedbacks' converged states.
[0026] FIG. 4 illustrates a further embodiment of a coarse channel
estimation unit, generally designated 400. A received sequence of
symbols along an input path 402 feeds respective first and second
unit branches 404 and 406. The first unit branch 404 includes an
optional linear equalizer 408 similar to that described with
respect to FIG. 3, with respective main and first precursor paths
410 and 412, delay element 414, and multipliers 416 and 418 to
receive respective main and first pre-cursor tap values a0 and al.
The multipliers feed their respective outputs to an input summer
420. A first slicer 422 couples to the output of the input summer
420 to generate decisions on the state of each received symbol.
[0027] Further referring to FIG. 4, the output of the first slicer
422 provides an input to a plurality of feedback paths similar to
the feedback paths described with respect to FIG. 3. In this
regard, the output of the first slicer 422 feeds a tapped delay
line of delay elements 424. Each element exhibits a delay of T,
corresponding to the symbol interval time, and has a tapped output
that is fed to a multiplier 426. The multipliers 426 receive
respective tap weightings W0-Wm determined by an LMS unit 430 to
apply to the delayed symbols from the delay elements 424. A
feedback summer 428 aggregates the compensation from each of the
multipliers 426, and feeds the aggregated output to the input
summer 420.
[0028] With continued reference to FIG. 4, the second unit branch
406 includes a second slicer 432 with a tapped delay line of delay
elements 434 and multipliers 436 that form a set of feedback paths.
The set of feedback paths of the second unit branch 406 generally
forms a replica of the first slicer 422 and its associated feedback
paths. An error summer 438 sums the received sequence of symbols
and the compensation from the second unit branch feedback paths and
generates an error signal for the LMS unit 430. The error signal
may be used for successive iterations in adjusting the tap weights
to form a coarse detected sequence of symbols, output from the
first slicer 422 along path 440.
[0029] FIG. 5 illustrates a flowchart of steps employed in one
embodiment of a method of operation to generate a coarse estimated
channel response. The method may be used with either of the coarse
channel response estimation units described with respect to FIGS. 3
and 4. At step 502, an input sequence of symbols is received at the
input summer. As each symbol is sequentially fed to the slicer, a
decision is made, at step 504, regarding the symbol state. In one
embodiment, the symbol state may be determined to be +1, 0, or -1.
The symbol states are fed to the plurality of feedback paths, at
step 506, which are selectively activated to apply compensation to
the input summer based on prior received symbols. The symbol states
are also acted on by an LMS algorithm, at step 508, which
determines the appropriate weightings to apply to the equalizer
taps associated with each of the feedback paths.
[0030] As explained above, each of the coarse channel estimation
unit embodiments 300 and 400 employ a given number of feedback
paths. The optimum number of paths to use may be determined by a
correlation-based thresholding method. A correlation matrix,
applied at step 510, used for the method may be defined as:
R x ~ x ~ = E { x ~ x ~ T } = E { x ~ ( t ) x ~ ( t ) } E { x ~ ( t
) x ~ ( t - 1 ) } E { x ~ ( t ) x ~ ( t - 2 ) } E { x ~ ( t ) x ~ (
t - 1 ) } E { x ~ ( t ) x ~ ( t ) } E { x ~ ( t ) x ~ ( t - 1 ) } E
{ x ~ ( t ) x ~ ( t - 2 ) } E { x ~ ( t ) x ~ ( t - 1 ) } E { x ~ (
t ) x ~ ( t ) } ##EQU00002##
where {tilde over (x)}(t) is the detected signal. The thresholding
is defined as:
E { x ~ ( t ) x ~ ( t - .tau. ) } < .delta. max , ( .tau. = 1 ,
, .tau. max ) ; ##EQU00003## .tau. = 1 .tau. max E { x ~ ( t ) x ~
( t - .tau. ) } .tau. max < .delta. aver ; ##EQU00003.2##
where .tau..sub.max is the maximum observing delay, .delta..sub.max
is the maximum threshold, and .delta..sub.aver is the average
threshold.
[0031] Once activated, compensation from the activated feedback
paths may then be applied to the input summer. For the embodiment
shown in FIG. 3, the feedback paths may be incrementally activated
in sequence. For the embodiment shown in FIG. 4, the feedback paths
may be activated simultaneously. After one or more iterations, the
output from the slicer may be provided as an estimated channel
response that, at step 512, may be further refined by the fine
channel estimation circuit. In some embodiments, if the two
thresholds .delta..sub.max and .delta..sub.aver are satisfied, the
output sequence with the current DFE structure is assumed to meet a
bit-error-rate (BER) criterion, and the LMS-CE driven by the symbol
sequence may be used as the initial channel estimation for the fine
channel estimation unit 210 (FIG. 2). Otherwise, the output
sequence of the slicer and the LMS-CE may be selected as the
optimum choice for the coarse channel estimation.
[0032] As shown at a high level in FIG. 2, the coarse channel
estimation generated by the coarse channel estimation unit 208 is
fed to the fine channel estimation unit 206 to obtain further
convergence of the channel impulse response.
[0033] FIG. 6 shows a flowchart of steps employed in one embodiment
of a method to generate a fine channel estimation based on the
coarse channel estimation of FIG. 5. At step 602, the fine channel
estimation unit receives the initial channel estimation from the
coarse channel estimation unit. As explained above, the fine
channel estimation unit includes an MLSE unit that employs Viterbi
coding techniques to more accurately determine the symbol sequence.
The MLSE also cooperates with an LMS-CE unit which may be separate
from or shared with the coarse estimation unit. At step 604, a
number of consecutive taps employed by the LMS-CE unit may be
activated. In one embodiment, the consecutive taps may be selected
based on maximum energy peak values above a predetermined
threshold. The number of selected taps also determines the number
of states used in the MLSE algorithm.
[0034] Further referring to FIG. 6, once the channel estimation
window of taps is selected, the fine channel estimating begins, at
step 606, by generating path metrics and branch metrics consistent
with Viterbi algorithm techniques. In one embodiment, a path metric
calculated in the MLSE may be represented by the expression:
.LAMBDA.(n)=.LAMBDA.(n-1)+.lamda.(n)
where .LAMBDA.(n) is the accumulative path metric from the start to
sample time n, and .lamda.(n) is a branch metric. The branch metric
may be expressed as:
.lamda.(n)=|y(n)-h.sub.MLSE.sup.T(n)x(c(n-1).fwdarw.c(n))|.sup.2
where c(n) represents state in a trellis diagram, and h.sub.MLSE is
the CSI for MLSE. With the path and branch metrics calculated for a
given sample time n, a comparison may be carried out, at step 608,
with the accumulative path metrics, and an optimum decoded symbol
determined. If the number of samples at the given sample time n is
less than a predetermined number of samples N, determined at 610,
then additional samples may be accumulated, at 612, and new branch
and path metrics calculated based on the new samples, at 606.
[0035] With continued reference to FIG. 6, once the number of
samples reaches the threshold N, the process may enter a tracking
mode, at 614. The tracking mode generally operates to detect
changes in the channel response, and adaptively update the tap
weights to compensate for any changes, at step 616. Following each
update, a determination may be made, at 618, as to whether new
channel metrics exceed a threshold level. If so, then additional
samples may be accumulated, at 620, to allow for further metric
generation and updating. If not, then the tracking process
terminates, and a new window of consecutive taps may be selected
for use by the MLSE, at step 604. Steps 606 to 616 may then be
iterated.
[0036] Typically, there are L'(L'<L) taps before the estimation
of the first actual channel tap in the channel estimation window
after the initial coarse channel estimation process. The MLSE also
introduces a trace back delay .tau.', which may be used to specify
the number of trellis branches (the number of symbols) that the
Viterbi algorithm uses to construct each trace back path. A trace
back depth influences the decoding accuracy and delay. In order to
refine the channel estimate based on the acquired channel estimate,
the LMS-CE may be updated according to the following
expressions:
e(n)=y(n-L'-.tau.')-h.sup.T(n){tilde over (y)}'(n)
{circumflex over (h)}(n+1)={circumflex over (h)}(n)+.mu.e(n){tilde
over (y)}'(n)
where {tilde over (y)}'(n) is the output of the MLSE.
[0037] FIG. 7 illustrates respective examples of coarse and fine
estimations of a single-bit channel impulse response (CIR) curves
using the circuits and methods described herein. Each channel
impulse response curve shows a main cursor h.sub.0 and various
post-cursors, such as h.sub.1 and h.sub.2, for comparison purposes.
A pre-cursor h.sub.-1 is also shown for situations where an
optional linear equalizer is employed for pre-cursor ISI
mitigation. Adjusting a given post-cursor tap in accordance with
any of the techniques described herein causes the corresponding
post-cursor to exhibit a different response characteristic. By
generating an accurate channel impulse response based on the
circuits and methods described above, sequences of symbols can be
accurately determined under challenging signaling conditions.
[0038] Those skilled in the art will appreciate the benefits and
advantages afforded by the embodiments described herein. For
example, the channel estimation circuitry and methods described
herein beneficially provide a more straightforward and less complex
way to estimate the channel response. By minimizing the complexity,
the circuits and methods described herein are more suitable for
implementation in systems such as 10GBASE-LRM.
[0039] In the foregoing specification, the present embodiments have
been described with reference to specific exemplary embodiments
thereof. It will, however, be evident that various modifications
and changes may be made thereto without departing from the scope of
the disclosure as set forth in the appended claims. The
specification and drawings are, accordingly, to be regarded in an
illustrative sense rather than a restrictive sense.
* * * * *